Vermont Blueprint for Health

Burlington Area Community Network Report Network Analysis and Team Based Care

Maurine Gilbert, contracted Community Facilitator July 2015

Objective Describe the network of organizations that has emerged in each Blueprint Health Service Area (HSA) to

support population and individual health, focusing on modes of collaboration and relationships between

organizations.

Background and Key Questions The Vermont Blueprint for Health is a state-led, nationally-recognized initiative transforming the way

primary care and comprehensive health services and delivered and paid for. The Blueprint encourages

the growth of regionally-based multi-disciplinary networks of health, social and economic service

providers. These networks are intended to bring a diverse group of service providers closer together, to

deliver more seamless and holistic care to the people of their regions. This study is the first step towards

answering key questions about the networks that are active in Blueprint communities: What role did

investment in core Community Health Teams have in seeding these larger networks? How are the

participating organizations connected to each other? How are these relationships maintained and

reinforced – how durable are they? What characteristics do the most successful networks share? And,

ultimately, what impact do that have on individual and population health?

Approach This study used a combination of network analysis, investigating connections between organizations,

and traditional polling methodology, addressing the experience of working together as a team.

Network Analysis Network analysis was the central methodology in this study, used for its ability to characterize and

quantify relationships in a complex system. Network analysis creates graphs that show the connections

between individuals or (as in this case) organizations. With these graphs and quantitative network data,

researchers and community members can explore the relationships that make up the network and start

to look for patterns as well as changes over time. Observations of network data and network graphs can

lead to smarter, better questions about how community-based teams coalesce and how they create

change.

The data used in this study are responses to a survey question that asked representatives of

organizations to report whether their organization interacted with other organizations in their area in

any (or all) of six ways, stated as follows:

1. “My organization sends referrals to this organization”

2. “My organization receives referrals from this organization”

3. “Our organizations have clients/patients in common”

4. “Our organizations share information about specific clients/patients”

5. “Our organizations share information about programs, services and/or policy”

6. “Our organizations share resources (e.g. joint funding, shared equipment, personnel or

facilities)”

Additionally, several questions were included in the study that were not intended for network analysis.

These included demographic questions and a set of questions about whether respondents perceived

their communities to be acting as teams.

Team Based Care In 2012 The Institute of Medicine (IOM) published the discussion paper “Core Principles & Values of

Effective Team-Based Health Care.” The Vermont Blueprint for Health embraces this paper’s model, of

how a team should function and feel, as a goal for both direct clinical care and multidisciplinary

community health improvement. The five hallmarks of effective team based care given by the IOM are

Shared Goals, Mutual Trust, Clear Roles, Effective Communication, and Measureable Processes and

Outcomes. In the FY2015 survey, respondents were asked to think about how all of the organizations

listed work together as group, and agree or disagree with statements about whether they exhibit each

of those hallmarks of team-based care.

List Development Over the course of the 2015 network survey, the list development methodology used for this study was

adjusted twice in response to findings from the research, which was conducted in waves. Each

adjustment pushed the network bounding towards greater consistency across HSAs and towards smaller

network membership lists and shorter survey instruments.

This HSA was included in the third wave of data collection, using the Core Network List Development

methodology. With this methodology, the network list was a core group of organizations similar to the

organizations represented in the area’s Unified Community Collaborative, as shown below. No

additional organizations were included.

Types of Organizations Included in Core Network Methodology

Community Health Team

Each Blueprint PCMH primary care practice

Known non-Blueprint primary care practices

FQHC dental clinic

Hospital

Hospital – Emergency Department

Hospital – Case Management/Social Work Department

Designated Mental Health Agency

“Hub” of Hub/Spoke Program

VNA

Area Agency on Aging

Designated Regional Housing Organization – SASH Program

State of VT – Agency of Human Services (AHS)

State of VT – Vermont Chronic Care Initiative (VCCI)

State of VT – Vermont Department of Health (VDH)

area United Way

Survey Participation Invitations Sent 51

Surveys Started 31

Response Rate 61%

Completed Surveys 26

Completion Rate 84%

Organizations Completed Survey

Alder Brook Family Health Y

Burlington Primary Care

Cathedral Square - SASH Program Y

Champlain Center for Natural Medicine

Champlain Valley Agency on Aging (CVAA) Y

Charlotte Health Center

Community Health Centers of Burlington Y

Community Health Centers of Burlington - Dental Clinic

Essex Pediatrics Y

Evergreen Family Health Y

Good Health Y

Hagan, Rinehart and Connolly Pediatricians Y

Howard Center Y

Howard Center - The Chittenden Clinic "Hub"

Mountain View Natural Medicine Y

Practice of Frank Landry, MD PLC

Practice of Gene Moore, MD

Practice of James Hebert, MD

Richmond Family Medicine

State of VT - Agency of Human Services (AHS) Y

State of VT - Vermont Chronic Care Initiative (VCCI) Y

State of VT - Vermont Department of Health (VDH) Y

Thomas Chittenden Health Center

Timber Lane North Pediatrics

Timber Lane Pediatrics Y

United Way of Chittenden County Y

University of Vermont Medical Center (UVMMC) Y

UVMMC - Adult Primary Care - Burlington Y

UVMMC - Adult Primary Care - Essex

UVMMC - Adult Primary Care - South Burlington

UVMMC - Adult Primary Care - Williston Y

UVMMC - Case Management and Social Work Department Y

UVMMC - Community Health Team Y

UVMMC - Emergency Department Y

UVMMC - Family Medicine - Colchester Y

UVMMC - Family Medicine - Hinesburg

UVMMC - Family Medicine - Milton Y

UVMMC - Family Medicine - South Burlington Y

UVMMC - Pediatric Primary Care - Burlington Y

UVMMC - Pediatric Primary Care - Williston

Visiting Nurse Association of Chittenden and Grande Isle Counties (VNA) Y

Winooski Family Health

Data Analysis Non-network data analysis was conducted in Survey Monkey and Excel.

Network analysis was conducted using Gephi. Data is input into Gephi in node lists and edge lists. Node

lists are lists of the names/labels of the organizations included in the study and a corresponding number.

Edge lists are lists of the connections between organizations. In this study each edge list represented all

the instances of a single type of connection (sharing resources, for instance) in a single HSA. The edge

lists began with an extract of data from Survey Monkey, a grid format recording each connection

between organizations. The grids were transformed in a series of steps into the edge lists, which code

connections in pairs of numbers giving the “Source” and “Target” of each connection. The edge lists

used in this study have been de-duplicated – in cases where multiple respondents answered on behalf

of a single organization the connection between that organization and any other organization will

appear only once per list. This choice was made to prevent over representing the role in the network of

organizations fielding multiple respondents.

Results

Network Analysis Glossary The following are brief definitions of network terminology that will be used throughout the Results

section.

Node

The “nodes” on these graphs are the dots that represent organizations

Edge

The “edges” on these graphs are the lines representing connections between organizations

(connections of any sort, whether they represent sharing information, resources, or referrals)

Centrality

Importance or prominence of an actor in a network

Betweenness Centrality

A measure of how often a given node appears on the shortest paths between pairs of nodes in

the network. Betweeness Centrality takes the entire network into consideration when

calculating a score for an individual node, and is therefore considered one of the most powerful

centrality measures.

Average Degree

The average number of edges connected to each node in the network

Average Shortest Path Length

The average number of edges on the shortest path between each pair of nodes in the network

Graph Density

The proportion of all possible connections (represented as edges) that are present

Modularity

A measure of how readily a network decomposes into modular communities or sub-networks.

The modularity numbers given here are based on the modularity function used in the Gephi

software program (there are many other “modularity” or “community detection” functions that

may be used in network analysis.

Network Maps See Appendix A for the Network Maps

Network Statistics

Common Patients

Info – Patients

Info – Programs Resources Referrals

Full Network

Avg. Degree 10.143 7.095 5.167 3.571 10.19 16.048

Avg. Weighted Degree 10.413 7.095 5.167 3.571 12.786 38.762

Network Diameter 3 3 3 4 3 2

Graph Density 0.247 0.173 0.126 0.087 0.249 0.391

Modularity 0.092 0.142 0.188 0.351 0.095 0.083

Avg. Clustering Coefficient 0.619 0.702 0.702 0.619 0.693 0.745

Avg. Path Length 1.517 1.715 1.858 1.877 1.781 1.613

Organization Statistics

Organizations Ranked by Betweeness Centrality

1 University of Vermont Medical Center (UVMMC)

2 UVMMC – Community Health Team

3 Visiting Nurse Association of Chittenden and Grande Isle Counties (VNA)

4 UVMMC – Case Management and Social Work Department

5 Mountain View Natural Medicine

Organizations with Highest In-Degree

University of Vermont Medical Center (UVMMC) 42

UVMMC – Community Health Team 41

Visiting Nurse Association of Chittenden and Grande Isle Counties (VNA) 40

UVMMC – Case Management and Social Work Department 40

State of VT – Vermont Chronic Care Initiative (VCCI) 29

Team-Based Care

Observations and Opportunities The following are the researcher’s observations of the network graphs and team based care results, and

related questions. Additional observations, questions, and ideas for improving network relationships

and effectiveness will be solicited when these findings are presented in the community.

In the full network:

o All of the most central organizations appear in the blue network neighborhood – these

include UVMMC and its Community Health Team, Emergency Department, and Case

Management and Social Work Department, as well as the VNA and Mountain View Natural

Medicine.

o State services are all clustered in the red network neighborhood.

o The green network neighborhood may coalesce around the high-needs populations some of

the organizations in that neighborhood serve. The Howard Center, its Chittenden Clinic

“Hub,” and the Community Health Centers of Burlington likely have significant overlap in

clientele.

The sub-network of information sharing about specific clients/patients is dominated by UVMMC and

the VNA – they are highly central as well as closely linked to each other. What systems or practices

have these organizations put in place to facilitate this information sharing, do they feel effective to

those utilizing them, and if so can they inform other HSAs? Note that these same organizations are

similarly positioned in the referrals network.

The Team Based Care measures show little change since the last survey, with the exception of a

drop of 10% in agreement that the community has established shared goals, and a 5% increase in

agreement that organizations trust each other.

72% 72%

61%54%

37%

62%

77%

64%

52%

40%

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Shared Goals Mutual Trust EffectiveCommunication

Clear Roles Measureable Processesand Outcomes

% of Respondents who "Agree" or "Strongly Agree" That Their Community Exhibits the Given Team Characteristics

FY14 FY15

Appendix ABurlington Network Maps

Burlington Common Clients NetworkOur organizations have clients/patients in commonNode color shows DegreeNode size shows Betweeness Centrality

Burlington Info-Patients NetworkOur organizations share information about specific patients/clientsNode color shows DegreeNode size shows Betweeness Centrality

Burlington Info-Programs NetworkOur organizations share information about programs, services and/or policyNode color shows DegreeNode size shows Betweeness Centrality

Burlington Resources NetworkOur organizations share resources (e.g. joint funding, shared equipment, personnel or facilities)Node color shows DegreeNode size shows Betweeness Centrality

Burlington Referrals NetworkMy organization sends referrals to this organization + My organization receives referrals from this organizationNode color shows DegreeNode size shows Betweeness Centrality

Burlington Full NetworkNode color shows Network NeighborhoodNode size shows Betweeness Centrality